Kandinsky 6.0 Video open-sourced under MIT, vLLM adds day-0 inference support
vllm_project · x · 2026-10-06
Kandinsky 6.0 Video is open-sourced under MIT, offering Lite (3B) and Pro (29B) diffusion models that generate 5-second video clips with synchronized 44kHz audio and lip-sync, in both text-to-audio-video (T2AV) and image-to-audio-video (TI2AV) modes. A plugin-in super-resolution model raises output to Full-HD (1920×1080).
vLLM-Omni announced day-0 support, so the models can run inference on vLLM from launch day. The repo also ships a ComfyUI plugin and quick-start scripts; requires an NVIDIA GPU and Python 3.13+, with the Pro Distill 5s weights downloaded automatically on first run.
Related event: Kandinsky 6.0 Video Open-Sources Audio-Video Generation Under MIT License(4 posts)→
More from Multimodal
- Google rolls out Nano Banana 2.1 with 4K output at $0.076 per image — testingcatalog · 2026-10-06
- 85mm Editorial Portrait Prompt Template That Preserves Facial Identity — aziz4ai · 2026-10-06
- Solaya turns a 3-minute iPhone scan into a photorealistic 3D digital twin in under an hour — willeastcott · 2026-10-06
- Dev makes promo video with Opus and fframes, skipping hours of After Effects work — kevinkern · 2026-10-06
- Viral prompt recipe makes GPT image models shoot 'unpublished' photojournalism — techhalla · 2026-10-06
- Fun Seedance video demo shared with the exact prompts to recreate it — techhalla · 2026-10-06